Introduction

• Fire Exits
• Toilet Locations
• Food and Coffee: always in the Resource Area
• Mugs for beverages
  • Please grab one in the morning.
  • It's yours for the day for all drinks.
  • At end of day, please wash in kitchen beside Resource Area
  • This helps us conserve on disposable cups and glasses.

• What questions are best answered using tags?
• What questions are you using tags to answer?

• Study species: Ziphius cavirostris
• Tag: DTAG3

Controlled Exposure Experiments (CEEs)

with Mid-frequency Active Sonar (MFAS)

• Can the question be answered with existing data?
• Are simulations a viable alternative?
• Is the tag suitable?
  • sensors, sampling rates
  • accuracy
  • size
  • tagging duration

• Lack of existing data for analysis or simulation
• Was tag suitable?
  • What sound did whales receive?
  • Do dive behavior, body movement or foraging behavior indicate response?

• Visual observations
• Acoustic tracking
• Visual and radar tracking of vessels in the area
• Photos and video
• Genetics (biopsy, skin sample)
• Life history of tagged whale (photo-id studies)
• Oceanographic measurements (CTD, current, bathymetry)

• Storage, quality control, archiving (paper? the dreaded spreadsheets?)
• Accounting for effort
• Collaboration with others in the study area, data sharing
• Integrating data with different time scales, varying accuracy

• How will you analyse the data to address your question?
• Who else can use the data?
• When and how will you share the data?
• What is “raw” data and is it what you should share? <!– - Does the value of the science merit the abuse? –> <!–
  –> <!– –> <!– –> <!–
  (photo from Dietz et al. 2007, doi.org/10.1186/1472-6785-7-14) –> <!– –> <!–
  –> <!– Note: maybe move this to the very end (earn your coffee examples) and include here the “running example” – cost (in many senses) per data point? –>

• Short-term (day) to Long-term (year)
• Attachment: suction cups, bolts, straps, glue…
• Sensors: Geo position, orientation, depth/altitude, sound, video, heart-rate, temperature…
• Archival vs. Telemetry

• Availability of tags & analysis tools
• “Ease” of attachment (for you and animal)
• Record duration vs. data density
• “Ease” of data retrieval

• What question are you trying to answer?
• Is your tag appropriate?
• We've chosen… and you (may have chosen)…

• Typically include:
  • Sensors sampled multiple times per second
  • Accelerometers to measure body posture and sudden movements or changes in speed
  • Magnetometers to measure direction of travel
  • Pressure/altitude sensors to measure dive depth in aquatic/marine species or flight altitude in birds

• Challenges
  • “Big” datasets
  • Varying sensor sampling rate, accuracy, synchronization
  • Interdisciplinary: Engineering, Biology, Physics, Math/Stat…

• Motivation for tool kit and workshop
  • Make tools freely and widely accessible
  • Provide better documentation
  • Build connections across tagging community
  • Work toward standardization across tag types, species, …
  • Provide on-ramp for new users and idea sharing for experts

• Tag data are full of surprises!
• Creativity is essential to make the most of your data
• Goal is empowering researchers to explore, not automation of all data processing

• Matlab
  • Pros: Broad user base, flexible, fast at matrix calculations
  • Cons: Expensive, steep learning curve
• Octave
  • Pros: Free & open source, “almost matlab” now
  • Cons: Basic user interface, steep learning curve
• R/RStudio
  • Pros: Free & open source, pro stats standard, tutorials/resources
  • Cons: Slower (?), steep learning curve

Check out some online learning resources gathered at:

https://github.com/stacyderuiter/TagTools/wiki/Software

Learn and contribute at

http://www.animaltags.org

It's a wiki - join us!

• In Matlab/Octave, input arguments to functions are not named
• User must enter arguments in a set order, and software interprets accordingly depending on how many inputs there were
• If you want to leave out one input and specify a later one, you may have to enter “[]” for the earlier input

• In R, the user can enter un-named input arguments; if so they must be in the “expected” order (just like in Matlab/Octave)
• In R, the safe/flexible option is to name inputs to a function. If inputs are named, the order they are specified in doesn't matter.
  • Imagine you want to plot p as a function of t. You could use:
    • plot(t,p)
    • plot(x=t, y=p)
    • plot(y=p, x=t)

• Some Matlab/Octave tag tools use specific input arguments differently depending on how many inputs are provided (and what kind of objects they are).
• Example: plott.m (Matlab/Octave) can have from one to many inputs. The second one is interpreted as:
  • a sampling rate (fsx),
  • an indicator for whether to reverse the y-axis scale ®, or
  • a second sensor data input
  • …depending on how many total inputs there are.

• plott.R has 11 different named inputs
  • only one of the 11 is absolutely required
  • user decides which inputs to supply and uses names
  • ex.: plott(X, r=TRUE, interactive=TRUE)

• Tag tools can read in:
  • Raw data from several tag types
  • archival format netCDF files (<– our focus)
• NetCDF files (.nc) must be stored locally before reading.
• Assume we have a file stored at C:\data\testset1.nc
• R:
  

testset1 <- load_nc('C:\data\testset1.nc')

• Matlab/Octave:

testset1 = load_nc('C:\data\testset1.nc') ;

Metadata has at least 2 sources:

• Gleaned automatically from tag data files
• User-provided

• Some metadata is automatically gleaned and stored in a csv when you use the tag tools to load raw data
• An interface is provided to add user-generated metadata to this machine-generated file.
• If the machine-generated file is 'auto.csv', in R or matlab:

metadata_editor('tagmetadata.html', 'auto.csv')

• Code from last slide will open a fillable form.
• Some fields are always auto-gleaned and don't appear on the form for users to fill in.
• If you use the form without a machine-generated csv input, these fields will all contain missing values in the resulting csv.

There's currently no “standard” format for tag data;

Raw datasets from most tags are not archive-ready.

netCDF format…

• Developed with NSF funds for geoscience community
• Is flexible, accomodates data and metadata fields
• Allows many user-defined fields
• Supported by well-defined standards and conventions
• May serve as a standad format for journals that require data publication
• “supports a view of data as a collection of self-describing, portable objects that can be accessed through a simple interface” (-NetCDF website)

Animal Telemetry Network (ATN) Data Assembly Center (DAC)

Animal Telemetry Network (ATN) Data Assembly Center (DAC)

• In R, all “tools” must be bundled in packages.
• To use a package, you must:
  • Install the package (one-time-only)
  • Load the package (every new R session)

Option A) Install “development version” from GitHub

#install.packages('devtools')
devtools::install_github(
  'stacyderuiter/TagTools/R/tagtools')

Option B) Install from your memory stick. In RStudio:

• Select the Packages tab in the lower right pane
• Click Install
• Choose Install from package archive file
• Click Browse and navigate to the R tagtools zip file on your memory stick
• Click Install!

• Matlab has access to all functions that are:
  • In directories that are on the Matlab Path
  • In the user's current directory (“Current Folder” in GUI)

• Unpack the Matlab tagtools zip file on your computer
• In Matlab, in the Home tab, click Set Path
• Click Add with Subfolders and then select the tagtools directory
• Click Save to save the modified path

• Octave has access to all functions that are:
  • In the user's current directory (“File Browser” in GUI)
  • In packages that have been
    • installed (one-time operation) and
    • loaded (every new Octave session)

• Save the matlab/octave tagtools directory to a convenient location on your computer
• Add that tagtools directory to Octave's path (change the path to match your computer):

path_tagtools = genpath('C:\tagtools\')
addpath(path_tagtools)
savepath

In Octave, you also need to manually install and load the packages:

• control
• netcdf
• signal
• io (for reading DTAG files)
• and possibly: statistics, mapping, ncarray, lssa,ltfat, database.

If you are online, you can simply run (for example):

pkg install netcdf 
pkg load netcdf

• You can customize Octave's start-up routine (autoload packages).
• See http://www.animaltags.org/doku.php?id=tagwiki:help:octave:autoload for details.

For specific tools, help is only a command away. Example: need help with the function plott()?

in Matlab/Octave:

help plott

in R:

?plott

• As possible, save data, code, and documents in subdirectories of one main project directory
• Document your work:
  • Save files rather than just running code in the console
  • Document your code with comments and explantions
  • Give variables names that are easy to understand
  • Prefer descriptive labels to numeric identifiers
• Consider Reproducible Research - code, documentation and report all in one document!
  • In R: knitr
  • In R: Rmarkdown
  • In Octave/Matlab… ?

(Stay tuned…Maybe after a coffee.)

(just a few!)

• Can you argue both ways?
  
  (photo from Dietz et al. 2007, doi.org/10.1186/1472-6785-7-14)